agentsclimarketplace
Sign in

Redaction and cleanup

Skill aneja5/forge-skills/skills/redaction-and-cleanup

Use when user says "redact", "clean up for sharing", "remove sensitive info", "prepare for external", when sharing .forge/ artifacts with investors or external reviewers, when sanitizing documents before external publication, or when pricing/credentials/strategy must be removed before sharing.From its SKILL.md

Install
npx -y skills add aneja5/forge-skills --skill redaction-and-cleanup

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 3 stars3 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

4.9 KB, ~1.1k tokens by cl100k_base, as published. Nobody here has run it

Redaction and Cleanup

Overview

Prepare .forge/ documents for external sharing by copying them to .forge/redacted/ and replacing sensitive information with consistent placeholders. Original files are NEVER modified. The redaction manifest documents every change for internal reference.

When to Use

  • Sharing documents with investors, advisors, or partners
  • Preparing materials for external review or cross-validation
  • Publishing case studies or documentation based on internal artifacts
  • Any situation where internal data must not leave the organization

When NOT to Use

  • Documents are already public — no redaction needed
  • User wants to delete files — that's not redaction
  • The sensitive data IS the deliverable (e.g., a security audit for the client)

Common Rationalizations

ThoughtReality
"I'll just manually remove the sensitive parts"Manual redaction misses things. Automated scan + manifest catches more
"It's fine, the recipient is trusted"Trust the person, redact the document. People forward things
"I'll edit the originals and restore later"You won't. Original files must be untouched — copy first, always
"Only pricing needs redacting"Credentials, internal metrics, employee names, strategy details — audit first
"A quick find-and-replace is enough"Inconsistent placeholders leak information through correlation

Red Flags

  • Original files modified in place (SAFETY VIOLATION — must copy first)
  • Placeholder values are inconsistent (same entity gets different placeholders)
  • Grep after redaction still finds sensitive terms
  • Manifest is missing (no record of what was changed)
  • Redaction categories not defined before scanning (ad-hoc misses patterns)

SAFETY RULE

NEVER modify original files in place. All redaction happens on copies in .forge/redacted/. Original files in .forge/ must remain untouched. If you are about to edit a file outside .forge/redacted/, STOP.

Core Process

Step 1: Define redaction categories

With the user, define what to redact and what to keep:

Redact (typical):

  • Pricing, margins, unit economics
  • API keys, credentials, connection strings
  • Internal strategy, competitive positioning details
  • Employee names, internal team structure
  • Specific customer names (unless public case study)
  • Internal metrics, revenue numbers

Keep (typical):

  • Public pricing from competitors
  • Architecture patterns and technology choices
  • General cost categories (without specific numbers)
  • Role titles (without names)
  • Publicly available data

Step 2: Define placeholder format

Use consistent, identifiable placeholders:

  • Company names: [COMPANY-A], [COMPANY-B] (consistent across all docs)
  • Prices: [PRICE-TIER-1], [PRICE-TIER-2]
  • People: [TEAM-MEMBER-1], [TEAM-MEMBER-2]
  • Metrics: [INTERNAL-METRIC]
  • Credentials: [CREDENTIAL-REDACTED]

Consistency rule: the same entity gets the same placeholder everywhere.

Step 3: Scan all files

Scan every file in .forge/ for matches against redaction categories. For each match, record:

  • File path
  • Line number
  • Original content (summary, not full text)
  • Redaction action (replace with which placeholder)

Step 4: Generate manifest

Write .forge/redaction-manifest.md:

  • Redaction categories defined
  • Placeholder mapping (which entity → which placeholder)
  • Per-file change list
  • Total changes count

Step 5: Copy and redact

  1. Create .forge/redacted/ directory
  2. Copy all files from .forge/ to .forge/redacted/ (preserving structure)
  3. Apply redactions to the copies only
  4. Do NOT copy .forge/redaction-manifest.md to redacted/ (it contains the mapping)

Step 6: Verify

Run grep on .forge/redacted/ for every term in the redaction categories. Zero matches required. If any remain, fix and re-verify.

Step 7: Sync check

If documents are updated later, the redacted copies are stale. Note in the manifest: "Redacted on [date]. If originals change, re-run redaction."

Output

  • .forge/redaction-manifest.md — what was redacted, placeholder mapping
  • .forge/redacted/* — cleaned copies of all documents

Verification

  • Original files in .forge/ are unchanged
  • All files copied to .forge/redacted/ before any edits
  • Placeholders are consistent (same entity → same placeholder across all files)
  • Grep confirms zero remaining matches for redacted terms in .forge/redacted/
  • Manifest documents every change with file and line reference
  • Redaction manifest is NOT copied to .forge/redacted/
  • .forge/redaction-manifest.md written

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Gives 0 of the 12 instructions most privacy data protection skills give in ~1.1k tokens

Counted across 120 of the 151 authors here whose files we hold, read 2026-09-06

  • Log every PHI access and modificationin 9 of 120, across 4 files
  • Sanitize stack traces before sending to error trackingin 9 of 120, across 4 files
  • Make the audit log insert-only and immutablein 9 of 120, across 4 files
  • Encrypt personal data at rest and in transitin 8 of 120, across 6 files
  • Enable row-level security on PHI and PII tablesin 8 of 120, across 3 files
  • Tag PHI and PII columns at schema levelin 8 of 120, across 3 files
  • Require authentication on all PHI endpointsin 7 of 120, across 2 files
  • Configure session timeoutin 7 of 120, across 2 files
  • Run the deployment checklist before every deploymentin 7 of 120, across 3 files
  • Log only opaque record IDs, not namesin 7 of 120, across 3 files
  • Document processing in records of processing activitiesin 7 of 120, across 4 files
  • Return generic errors and log details server-sidein 6 of 120, across 2 files

Said here and by no other author read

  • Copy originals to a redacted directory before any edits
  • Apply redactions only to copies
  • Define redaction categories with the user before scanning
  • Define a consistent placeholder format before redacting
  • Use the same placeholder for the same entity everywhere
  • Scan every file against the redaction categories

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.

Keep looking

Skills are one crate of 325,949. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.